Civil and Environmental Engineering


Date of this Version

Fall 12-6-2010


A DISSERTATION Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Doctor of Philosophy, Major: Engineering (Civil Engineering), Under the Supervision of Professor Laurence R. Rilett. Lincoln, Nebraska: December, 2010
Copyright 2010 Bhaven Naik


A core component within Advanced Traveler Information Systems is travel time information because it is easily understood and perceived by travelers. However, the suggested travel time information should be based not only on historical and real-time conditions, but also on forecasted or “unknown” future conditions.

Existing research has focused primarily on developing models that forecast point estimates of the mean travel time that are in close comparison to their respective field values. There has been limited research on any insight into the reliability or uncertainty margin that exists around the forecasted point estimate. As well, these researches have a limitation in that the methodologies suggested are applicable to datasets that are assumed to be independent and identically distributed. However, this is generally not the case for the readily and widely available Intelligent Transportation Systems’ data.

This dissertation identifies an approach that computes a forecasted travel time as well as an estimate of standard error (the basis for reliability measures in transportation) for highly nonlinear models. Additionally, the approach accounts for the dependent structure of a dataset. The approach is generic and could be applied to other estimation and prediction models as well as other traffic variables, such as flow and speed.

Whereas the ordinary bootstrap has been used previously for uncertainty modeling within the travel time prediction environment, it is ideal for dealing with data that are independent and identically distributed. The application of two other bootstrapping methods—the block bootstrap and the gapped bootstrap—is demonstrated. The block bootstrap is currently the best known method for implementing the bootstrap with dependent data. The gapped bootstrap is a recently developed technique that is uniquely suited for handling uncertainties in dependent data.

The results suggest that, for the datasets used in this dissertation, the gapped bootstrap adequately captures the dependent structure when compared to the ordinary and block bootstrap methods. As well, unlike the ordinary bootstrap which is suitable only for data that are independent, it appears the gapped bootstrap can adequately address uncertainties for both independent and dependent structured datasets.